#include "genxml/genX_pack.h"
#if GEN_GEN == 8
-static void
-emit_viewport_state(struct anv_cmd_buffer *cmd_buffer,
- uint32_t count, const VkViewport *viewports)
+void
+gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer)
{
+ uint32_t count = cmd_buffer->state.dynamic.viewport.count;
+ const VkViewport *viewports = cmd_buffer->state.dynamic.viewport.viewports;
struct anv_state sf_clip_state =
anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 64, 64);
struct anv_state cc_state =
}
anv_batch_emit(&cmd_buffer->batch,
- GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC),
- .CCViewportPointer = cc_state.offset);
+ GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC), cc) {
+ cc.CCViewportPointer = cc_state.offset;
+ }
anv_batch_emit(&cmd_buffer->batch,
- GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP),
- .SFClipViewportPointer = sf_clip_state.offset);
-}
-
-void
-gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer)
-{
- if (cmd_buffer->state.dynamic.viewport.count > 0) {
- emit_viewport_state(cmd_buffer, cmd_buffer->state.dynamic.viewport.count,
- cmd_buffer->state.dynamic.viewport.viewports);
- } else {
- /* If viewport count is 0, this is taken to mean "use the default" */
- emit_viewport_state(cmd_buffer, 1,
- &(VkViewport) {
- .x = 0.0f,
- .y = 0.0f,
- .width = cmd_buffer->state.framebuffer->width,
- .height = cmd_buffer->state.framebuffer->height,
- .minDepth = 0.0f,
- .maxDepth = 1.0f,
- });
+ GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP), clip) {
+ clip.SFClipViewportPointer = sf_clip_state.offset;
}
}
#endif
-#define emit_lri(batch, reg, imm) \
- anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_IMM), \
- .RegisterOffset = __anv_reg_num(reg), \
- .DataDWord = imm)
-
-void
-genX(cmd_buffer_config_l3)(struct anv_cmd_buffer *cmd_buffer, bool enable_slm)
-{
- /* References for GL state:
- *
- * - commits e307cfa..228d5a3
- * - src/mesa/drivers/dri/i965/gen7_l3_state.c
- */
-
- uint32_t l3cr_slm, l3cr_noslm;
- anv_pack_struct(&l3cr_noslm, GENX(L3CNTLREG),
- .URBAllocation = 48,
- .AllAllocation = 48);
- anv_pack_struct(&l3cr_slm, GENX(L3CNTLREG),
- .SLMEnable = 1,
- .URBAllocation = 16,
- .AllAllocation = 48);
- const uint32_t l3cr_val = enable_slm ? l3cr_slm : l3cr_noslm;
- bool changed = cmd_buffer->state.current_l3_config != l3cr_val;
-
- if (changed) {
- /* According to the hardware docs, the L3 partitioning can only be changed
- * while the pipeline is completely drained and the caches are flushed,
- * which involves a first PIPE_CONTROL flush which stalls the pipeline and
- * initiates invalidation of the relevant caches...
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .TextureCacheInvalidationEnable = true,
- .ConstantCacheInvalidationEnable = true,
- .InstructionCacheInvalidateEnable = true,
- .DCFlushEnable = true,
- .PostSyncOperation = NoWrite,
- .CommandStreamerStallEnable = true);
-
- /* ...followed by a second stalling flush which guarantees that
- * invalidation is complete when the L3 configuration registers are
- * modified.
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .DCFlushEnable = true,
- .PostSyncOperation = NoWrite,
- .CommandStreamerStallEnable = true);
-
- emit_lri(&cmd_buffer->batch, GENX(L3CNTLREG), l3cr_val);
- cmd_buffer->state.current_l3_config = l3cr_val;
- }
-}
-
static void
__emit_genx_sf_state(struct anv_cmd_buffer *cmd_buffer)
{
if (!cmd_buffer->device->info.has_llc)
anv_state_clflush(cc_state);
- anv_batch_emit(&cmd_buffer->batch,
- GENX(3DSTATE_CC_STATE_POINTERS),
- .ColorCalcStatePointer = cc_state.offset,
- .ColorCalcStatePointerValid = true);
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CC_STATE_POINTERS), ccp) {
+ ccp.ColorCalcStatePointer = cc_state.offset;
+ ccp.ColorCalcStatePointerValid = true;
+ }
}
if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
if (!cmd_buffer->device->info.has_llc)
anv_state_clflush(cc_state);
- anv_batch_emit(&cmd_buffer->batch,
- GEN9_3DSTATE_CC_STATE_POINTERS,
- .ColorCalcStatePointer = cc_state.offset,
- .ColorCalcStatePointerValid = true);
+ anv_batch_emit(&cmd_buffer->batch, GEN9_3DSTATE_CC_STATE_POINTERS, ccp) {
+ ccp.ColorCalcStatePointer = cc_state.offset;
+ ccp.ColorCalcStatePointerValid = true;
+ }
}
if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
ANV_CMD_DIRTY_INDEX_BUFFER)) {
- anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_VF),
- .IndexedDrawCutIndexEnable = pipeline->primitive_restart,
- .CutIndex = cmd_buffer->state.restart_index,
- );
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_VF), vf) {
+ vf.IndexedDrawCutIndexEnable = pipeline->primitive_restart;
+ vf.CutIndex = cmd_buffer->state.restart_index;
+ }
}
cmd_buffer->state.dirty = 0;
cmd_buffer->state.restart_index = restart_index_for_type[indexType];
- anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_INDEX_BUFFER),
- .IndexFormat = vk_to_gen_index_type[indexType],
- .MemoryObjectControlState = GENX(MOCS),
- .BufferStartingAddress = { buffer->bo, buffer->offset + offset },
- .BufferSize = buffer->size - offset);
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_INDEX_BUFFER), ib) {
+ ib.IndexFormat = vk_to_gen_index_type[indexType];
+ ib.MemoryObjectControlState = GENX(MOCS);
+ ib.BufferStartingAddress =
+ (struct anv_address) { buffer->bo, buffer->offset + offset };
+ ib.BufferSize = buffer->size - offset;
+ }
cmd_buffer->state.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER;
}
const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline);
const struct brw_stage_prog_data *prog_data = &cs_prog_data->base;
- unsigned local_id_dwords = cs_prog_data->local_invocation_id_regs * 8;
- unsigned push_constant_data_size =
- (prog_data->nr_params + local_id_dwords) * 4;
- unsigned reg_aligned_constant_size = ALIGN(push_constant_data_size, 32);
- unsigned push_constant_regs = reg_aligned_constant_size / 32;
-
if (push_state.alloc_size) {
- anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_CURBE_LOAD),
- .CURBETotalDataLength = push_state.alloc_size,
- .CURBEDataStartAddress = push_state.offset);
+ anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_CURBE_LOAD), curbe) {
+ curbe.CURBETotalDataLength = push_state.alloc_size;
+ curbe.CURBEDataStartAddress = push_state.offset;
+ }
}
assert(prog_data->total_shared <= 64 * 1024);
.BindingTableEntryCount = 0,
.SamplerStatePointer = samplers.offset,
.SamplerCount = 0,
- .ConstantIndirectURBEntryReadLength = push_constant_regs,
+ .ConstantIndirectURBEntryReadLength =
+ cs_prog_data->push.per_thread.regs,
.ConstantURBEntryReadOffset = 0,
.BarrierEnable = cs_prog_data->uses_barrier,
.SharedLocalMemorySize = slm_size,
.NumberofThreadsinGPGPUThreadGroup =
- pipeline->cs_thread_width_max);
+ cs_prog_data->threads,
+ .CrossThreadConstantDataReadLength =
+ cs_prog_data->push.cross_thread.regs);
uint32_t size = GENX(INTERFACE_DESCRIPTOR_DATA_length) * sizeof(uint32_t);
- anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_INTERFACE_DESCRIPTOR_LOAD),
- .InterfaceDescriptorTotalLength = size,
- .InterfaceDescriptorDataStartAddress = state.offset);
+ anv_batch_emit(&cmd_buffer->batch,
+ GENX(MEDIA_INTERFACE_DESCRIPTOR_LOAD), mid) {
+ mid.InterfaceDescriptorTotalLength = size;
+ mid.InterfaceDescriptorDataStartAddress = state.offset;
+ }
return VK_SUCCESS;
}
genX(cmd_buffer_flush_compute_state)(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
- const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline);
- VkResult result;
+ MAYBE_UNUSED VkResult result;
assert(pipeline->active_stages == VK_SHADER_STAGE_COMPUTE_BIT);
- bool needs_slm = cs_prog_data->base.total_shared > 0;
- genX(cmd_buffer_config_l3)(cmd_buffer, needs_slm);
+ genX(cmd_buffer_config_l3)(cmd_buffer, pipeline);
genX(flush_pipeline_select_gpgpu)(cmd_buffer);
}
cmd_buffer->state.compute_dirty = 0;
+
+ genX(cmd_buffer_apply_pipe_flushes)(cmd_buffer);
}
void genX(CmdSetEvent)(
ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_event, event, _event);
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .DestinationAddressType = DAT_PPGTT,
- .PostSyncOperation = WriteImmediateData,
- .Address = {
- &cmd_buffer->device->dynamic_state_block_pool.bo,
- event->state.offset
- },
- .ImmediateData = VK_EVENT_SET);
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ pc.DestinationAddressType = DAT_PPGTT,
+ pc.PostSyncOperation = WriteImmediateData,
+ pc.Address = (struct anv_address) {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ };
+ pc.ImmediateData = VK_EVENT_SET;
+ }
}
void genX(CmdResetEvent)(
ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_event, event, _event);
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
- .DestinationAddressType = DAT_PPGTT,
- .PostSyncOperation = WriteImmediateData,
- .Address = {
- &cmd_buffer->device->dynamic_state_block_pool.bo,
- event->state.offset
- },
- .ImmediateData = VK_EVENT_RESET);
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ pc.DestinationAddressType = DAT_PPGTT;
+ pc.PostSyncOperation = WriteImmediateData;
+ pc.Address = (struct anv_address) {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ };
+ pc.ImmediateData = VK_EVENT_RESET;
+ }
}
void genX(CmdWaitEvents)(
for (uint32_t i = 0; i < eventCount; i++) {
ANV_FROM_HANDLE(anv_event, event, pEvents[i]);
- anv_batch_emit(&cmd_buffer->batch, GENX(MI_SEMAPHORE_WAIT),
- .WaitMode = PollingMode,
- .CompareOperation = COMPARE_SAD_EQUAL_SDD,
- .SemaphoreDataDword = VK_EVENT_SET,
- .SemaphoreAddress = {
- &cmd_buffer->device->dynamic_state_block_pool.bo,
- event->state.offset
- });
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_SEMAPHORE_WAIT), sem) {
+ sem.WaitMode = PollingMode,
+ sem.CompareOperation = COMPARE_SAD_EQUAL_SDD,
+ sem.SemaphoreDataDword = VK_EVENT_SET,
+ sem.SemaphoreAddress = (struct anv_address) {
+ &cmd_buffer->device->dynamic_state_block_pool.bo,
+ event->state.offset
+ };
+ }
}
genX(CmdPipelineBarrier)(commandBuffer, srcStageMask, destStageMask,